基于区域信息及改进综合学习粒子群优化的SAR图像配准方法

发布时间：2018-05-29 06:22

本文选题：SAR图像 + 显著区域　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：合成孔径雷达(SAR)采用微波主动成像技术,具有全天候、全天时、对地表和云层的穿透、高分辨率,侧视成像等特点,其成像条件不受天气或者能见度的影响,在遥感任务中得到了越来越广泛的应用。由于同一SAR系统在不同时期、不同视角经过同一区域时存在一定的几何形变,该问题的解决是图像融合、变化检测、地物分类等后续SAR图像处理的先决条件,因此,SAR图像配准是图像处理领域一个非常重要的研究课题。通过对现有配准的深入学习和研究,本文从区域角度探索解决不同时相SAR图像的配准问题,并做了相关仿真实验。本文的主要工作有以下内容:1.从显著区域分割的角度在浮动图像中提取区域,并用椭圆化的区域作为模板进行在参考图像中进行模板匹配,通过对浮动图像和参考图像构建小波金字塔,在不同分辨率上利用综合学习粒子群算法对匹配区域进行粗搜索和精搜索,在最顶层确定最佳匹配区域对,在其余层利用浮动图像中对应的该最佳区域继续进行区域搜索,最后利用匹配椭圆区域的中心点、长轴端点、短轴端点计算最终变换参数。对Radarsat-2卫星分别于2008和2009年所拍摄的黄河入海口的图像进行了实验,实验结果验证了本方法可以对SAR图像进行快速有效配准。2.提出采用基于灰度直方图众数排序的方法进行区域匹配,可以在两个区域分割效果并不是很一致的情况下很好地度量其相似性;对子区域图像对提出采用比值图像熵为相似度量函数,搜索算法为提出的改进后综合学习粒子群算法。所提出的配准方法,用子图像对的配准代替原图像对的配准,有效提高了配准效率,之后借助特征点在子图像对和原图像对中的不同坐标,将子图像对的配准参数转化为原图像对的配准参数。对Radarsat-2卫星分别于2008和2009年所拍摄的黄河入海口的图像进行了实验,实验结果验证了本方法的有效性。
[Abstract]:The synthetic Aperture Radar (SAR) uses active microwave imaging technology. It has the characteristics of all-weather, all-day penetration of the surface and clouds, high resolution, side view imaging, etc. The imaging conditions are not affected by weather or visibility. It has been used more and more widely in remote sensing mission. Due to the geometric deformation of the same SAR system in different periods and different angles of view passing through the same region, the solution to this problem is a prerequisite for subsequent SAR image processing, such as image fusion, change detection, ground object classification and so on. Therefore, SAR image registration is a very important research topic in the field of image processing. Through the in-depth study and research on the existing registration, this paper explores and solves the registration problem of different phase SAR images from the regional point of view, and makes relevant simulation experiments. The main work of this paper is as follows: 1. The region is extracted from the floating image from the angle of significant region segmentation, and the elliptical region is used as template to match the template in the reference image, and the wavelet pyramid is constructed by floating image and reference image. At different resolution, the matching region is searched rough and fine by comprehensive learning particle swarm optimization algorithm, the best matching region pair is determined at the top level, and the region search is continued in the other layers using the corresponding region in the floating image. Finally, the final transformation parameters are calculated by matching the center point, long axis end point and short axis end point of the elliptic region. Experiments were carried out on the Yellow River entrance images taken by Radarsat-2 satellite in 2008 and 2009 respectively. The experimental results show that the proposed method can be used to quickly and effectively register the SAR images. This paper proposes a method of region matching based on gray histogram mode ranking, which can measure the similarity of the two regions when the segmentation effect is not very consistent. The ratio image entropy is used as similarity measure function for sub-region image pairs and the improved comprehensive learning particle swarm optimization algorithm is proposed for searching. In the proposed registration method, the registration of the original image pair is replaced by the registration of the sub-image pair, and the registration efficiency is effectively improved, and then the different coordinates of the sub-image pair and the original image pair are obtained with the help of the feature points. The registration parameters of sub-image pairs are transformed into the registration parameters of the original image pairs. Experiments were carried out on the Yellow River entrance images taken by Radarsat-2 satellite in 2008 and 2009 respectively. The experimental results show that the proposed method is effective.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】